Year: 2022
Type: dissertation
Abstract: Perinatal medicine has drawn increasing attention by the neuroscientific research community, as an early detection of most pathologies leads to more effective therapies and treatments. Monitoring fetal and neonatal brain development through brain imaging, in particular, helps studying and identifying abnormalities that commonly take place at a pre-natal stage and have effects on post-natal development. The fast-developing perinatal brain poses several challenges that the existing techniques for adult brain analysis are not able to overcome, leading to the development of specific medical image analysis techniques and methodologies. Many of these challenges, though, remain unsolved. The purpose of this thesis is to present an automatic perinatal pipeline for segmentation and analysis of both pre- and post-natal magnetic resonance imaging (MRI) of the brain, and investigate in and ex utero brain development under specific abnormal conditions. The key contributions of this thesis are twofold: in the first part we present an automatic pipeline for fetal and neonatal segmentation and cortical mesh extraction. We propose two newly constructed temporal and multi-subject atlases, and present their application within the pipeline for atlas-based segmentation, based on novel registration methods. We extract the brain cortical mesh and compute morphometric descriptors of cortical folding, namely: mean curvature, local gyrification index, sulcal depth, and cortical thickness. We apply the pipeline to fetal and neonatal images and compare segmentation accuracy to state-of-the-art methods and ground truth. Our results show that the introduction of the new templates together with our segmentation strategy leads to accurate results when compared to expert annotations, as well as better performances when compared to available methodologies. In the second part of the thesis, we present a longitudinal study, using a dataset of 30 subjects (15 healthy controls and 15 subjects diagnosed with ventriculomegaly (VM)), with structural MRI acquired in and ex utero for each subject. We investigated the impact of fetal VM on cortical development from a longitudinal perspective, from fetal to neonatal stage. Particularly, we studied the relationship of ventricular enlargement with both volumetric features and a multifaceted set of cortical morphometric features including mean curvature, local gyrification index, sulcal depth, and cortical thickness. Our results show significant effects of VM on both volumetric and morphometric descriptors of cortical development in specific areas of the brain including the occipital, parietal and frontal lobes. This study shows the potential of the developed pipeline to perform innovative neuroimaging studies.
Author: Urru, Andrea
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7008558316
